Study: 'AI slop' accusations are social gatekeeping, not real detection

A new study from researchers Jason Miklian and John E. Katsos examined the explosive growth of accusations of AI use in online discourse. Analyzing 25 million comments from Reddit and Hacker News between 2023 and 2026, they found that pejorative labels like 'AI slop' increased over tenfold on both platforms. The term 'AI slop' now dominates, making up 94% of all negative mentions, while older inauthenticity terms like 'shill' or 'astroturf' did not see a similar rise. The shift reflects a fast-growing trend of branding any suspicious or seemingly inauthentic prose as AI-generated.

The key surprise came from a matched-control test comparing accused and non-accused parent comments. The test found that prose features that statistically distinguish AI from human text do not predict which human text actually gets accused as AI. Instead, the accusations serve as social gatekeeping of perceived authenticity — a way for commenters to perform in-group signaling and protest structural changes in online writing. The researchers extend signaling theory, showing that substitute signals (like accusing 'AI slop') can grow even when inaccurate if the underlying detection problem cannot be solved at the non-expert level. This means detection technology cannot resolve the dynamic because the social function of accusations is increasingly about policing community boundaries rather than identifying actual AI-generated content.